Have you considered adding a slight cost for submitting articles? (perhaps 2-5 points of karma?) There's a reward for submitting anything first that other people are likely to submit, but no cost, so the new page is often clogged with industry buzz that drowns out more substantial submissions. Many excellent article fall off within an hour, and then they can't get reposted.
Having more articles on the front page that aren't based on a scramble for duplicate-submission karma would probably improve the overall discussion threads, too; those posts tend to draw a lot of shallow comments.
It would also help with spam. Win win win.
I also wonder what percentage of upvotes for submissions comes just from duplicate submissions - maybe those should be counted differently (or not at all)? If the front page is already full of threads about some news about Apple (or whatever), being first to submit a redundant (but distinct) post is disproportionately rewarded, yet reduces the signal/noise ratio even further.
The problem with this is it makes every submission into a karma game. If submitting costs karma, it will cause more people to think "will this earn enough points?" before submitting links.
... but that's not what you want people to think. You want them to think "is this interesting and on-topic for HN?" They're not the same question, as much as the former tries to approximate the latter.
If people focus too much on points, you're likely to get more groupthink and more industry buzz (since that stuff always gets a gazillion upvotes) instead of interesting articles.
Certainly, people think about link karma already, but I think this would make it worse. Creating a cost will cause even people who don't think about karma to think about it, even for just a moment -- which is probably not what you want.
You want them to think "is this interesting and on-topic for HN?"
My earlier proposal (http://news.ycombinator.com/item?id=2242453 ): deduct karma from submitter and _anyone who upvoted_ on submissions that were successfully flagged. That should get people to think the way you mention not only when submitting, but when upvoting as well.
Improve the submissions and comments will naturally improve -- submissions that are one-sided/gossipy/demagogic tend to attract similar comments. Whereas submissions full of technical content are often rich with links to further reading, contrasting viewpoints with cited evidence, etc. (Disclaimer: observations in this paragraph may be subject to fundamental attribution error).
I think there would need to be a little more to associate the action with the punishment. Since the point deduction would be pretty delayed from the voting, voters might not recognize which votes were poorly placed, and wouldn't learn to place better votes.
I wonder about that, too, but think it'd be worth experimenting. The cost should probably be small.
I find myself wondering, "if I submit this now, will enough people be awake/etc. to notice it before it spills off the page?" That has deterred me from submitting more interesting content than costing 2 or 3 points ever would.
Also, a lot of those upvotes come from duplicate submissions. I don't have detailed info, but probably a significant percent, likely enough to lift them out of the new feed and onto the front page.
Or, maybe duplicate submissions shouldn't give the original an automatic upvote? I don't know.
Or, maybe duplicate submissions shouldn't give the original an automatic upvote?
Why shouldn't they? The fact that someone else indepedently thought the article was good to submit seems to be at least as strong a sign of interest as a click of an arror.
The problem here is that it doesn't distinguish between these two cases:
1. Multiple people independently submitted a content they found, because they found it interesting. This is (generally) good. Maybe it's not the community's cup of tea, and gradually fades away, but that's fine.
2. Somebody notices some fresh, overtly linkbait-ish article. They post it ASAP, hoping that fifty other people will be slower on the draw, netting them fifty upvotes. (Those people may be doing it for the same reason.) Either way: fluff gets submitted. Maybe it has fifty upvotes, maybe there are fifty others that just crowded out something really fascinating. This sets up a feedback loop that rewards posting things because they're linkbait, not thought-provoking.
I suspect that rewarding duplicate submissions too much tips the scales towards rewarding submitting linkbait. Looking at the new page basically ever backs this up. I don't know if adding a cost to submissions, reducing duplicate-submission karma, penalizing everybody involved for flagged->deleted articles, or what will help, but that seems like the root of the problem, so let's talk about it!
If you don't think a story will get 2-5 up votes, then you probably shouldn't submit it.
Unless a link is linkbait attention-whoring or breaking news, it's basically luck whether it gets that first upvote in its first few minutes to cause it to hit the front page. If you patrol "new" often, you'll see this constantly: tons of things in "new" are relevant articles, but still have no upvotes.
This problem only gets worse as HN grows in size.
Now, obviously, I have >4k karma and I shouldn't care at all about such a small karma cost. But the psychological effect exists regardless.
Timing can play such a critical role in an article gaining traction, that a lack of votes is not necessarily indicative of a lack of quality - e.g. right now, there is probably a worthy article on the new page that someone actively using no-procrast won't see because it is in competition with this announcement from PG.
Crowdbooster is a YC company right? They should totally implement a feature for the best time to submit an HN article like they do currently do with the best time to tweet.
If you just polled new submissions and cataloged the number of upvotes over time, you could establish the best time to submit an article. If you want to get fancy, you could add some sort of categorization to the metric and break it down by category (like: "articles about node.js seem to get the most attention when submitted Tuesdays at 3:00pm")
>Aside, I like not showing the points on articles.
I'm actually surprised at how much I like not showing points at all. I generally pride myself on carefully considering people's arguments and voting accordingly (if at all), but I've found myself reading comments much more carefully. It's too soon to tell what the effect will be in aggregate, but I my initial reaction is to hope the change is permanent.
The role of timing is interesting. It makes me wonder if a time-of-day-and-week based weighting could help out to reduce the need for posters to be so concerned about when they are posting.
What if the front page had a few random picks from the "new" page to give a little random extra exposure to undiscovered stories? They'd be different for each user...
Adding to your comment, it could also encourage group-think: "I'll only submit what I know others find cool." Note: I'm saying this as someone who submits stuff I think is very cool but doesn't get voted up very often.
I'm pretty surprised by the amount of fundamental attribution error displayed in this thread.
If each of us is so confident of our own abilities to rate articles impartially, be objective, etc., why are we putting so much thought into how to ensure that we can keep each other objective? If I believe that I can do it, why can't I believe that you can do it too?
I like the way this discussion is going. It is strange how karma could encourage people to act in a counter productive way.
1. How about not displaying the karma of new users for the first year?
2. The second insight is multiple postings from a news site. HN was good at recommending "hard to find" articles or esoterica. Where there are multiple people submitting the same link, it is probably not going to be hard to find or esoteric in the first place.
I don't get why a cumulative karma score even exists.
I like to see what other people think of my opinions and if I've touched a nerve. And I guess average comment score could conceivably be useful. But Total Karma just seems pointless. If the idea is to identify the oldtimers, then show their join date and be done with it.
The point to cumulative karma is to feed the rodent-like portion of our brain which likes getting slow, semi-random rewards. There's a tiny amount of satisfaction you get every time you look at your karma score and see that it's gone up since last time you looked at it... and a tiny amount of disappointment every time you see it go down.
Exactly. It's the same basic principle that Farmville uses for more nefarious purposes. It's a method of encouraging participation and engagement, which makes it very effective for cultivating a community of like-minded individuals.
It's probably an experiment in reputation. There was time when people would try to optimize their site's PageRank. The moment you start measuring things, magical emergent behaviors are created.
Even though it might not be the perfect way to estimate topicality for a comment/submission, karma is the best thing we have. As things are at the moment, groupthink is a huge problem no matter what rules you place on the system. You can't really stop it.
Everyone thinks about karma to some extent as it is, given that you can only do certain things with a certain amount of the stuff (downvoting, flagging, etc).
Even though it might not be the perfect way to estimate topicality for a comment/submission, karma is the best thing we have.
I don't think so. "Topicality" matches "lack of flagging", not "many upvotes". Or at least, it should -- there should be some standardized way, used by most HN members, of marking that they don't think a submission is good HN content.
This actually brings us to a very common problem on HN caused by the fact that you can't downvote links. It works like this:
Suppose you have two links, Link A and Link B. Link A is stupid TechCrunch linkbait, and Link B is an interesting scientific article. Link A is something everyone on the site understands, and 100 people upvote it. But in reality, suppose 2/3 of the people on HN actually would have downvoted it for being a terrible article -- but couldn't -- because there's no downvote.
Link B doesn't have as wide an appeal and gets only 20 upvotes, but nobody thinks it's a bad link. Yet despite 2/3 of the people on HN thinking that the TechCrunch article is terrible, Link A rates way higher than Link B.
While downvotes have their downsides, the fact that HN has no link downvotes basically guarantees that stupid linkbait industry buzz fills the top stories constantly. This is basically the HN equivalent of the Bikeshed problem.
It also gives people an incentive to submit the most linkbaity, overdramatized links possible, because the goal is not to submit good links: it's to submit links that get lots of upvotes. And when there's no downvotes, the best way to get upvotes is to get as much attention as humanly possible.
There are two types of articles that belong here on HN. The first type includes articles that speak for themselves. Everyone learns from them, but fell little need to comment unless they have something small to chime in. This type should have a low ratio of comments to upvotes.
However, the other type includes articles that provoke people and cause discussion. It doesn't just include contributing comments, but rather it contains debates and opposing opinions. These types of articles need a balance between a low and high ratio. However, the religious arguments you speak of are usually easy to find. These arguments are usually found in a deeply nested comment thread. Once you discard of fourth level comments and all the noise below, you should find that the ratio of comments is a pretty good signifier of the quality of the article.
It's very rare that there is a huge volume (>100 comments) of insight provoked by a single article, just because there's not usually that much insight to be had in the first place on that topic, and you can be sure that 100 separate comments didn't have it all on the same thread.
You usually can only get that much noise on a topic by poking people such that they NEED to correct some injustice done by the article. Sometimes good articles poke people like this, but in either case the comment thread ends up sucking.
I second this. It baffles me when I see an article with 100+ karma and 5 comments. How can something be so interesting and yet spawn so little discussion? It makes no sense.
As pg says, those are usually very good articles. I'd say the first impetus to comment is to correct, the second to add relevant detail. When the article doesn't need either, it's good.
There's a reverse problem too with complaining about karma farming eroding the SNR. I care about learning interesting things and couldn't care less about my karma on a website. I've only ever submitted two articles, and those were only after I read something that I found profoundly interesting that had just been posted (I read them right after getting an email alert that they were posted). For both, I also took the time to search HN to make sure they hadn't been submitted yet, before I hit the button. But the second one still drew complaints of being a duplicate based on prior submissions of a related previous Techchrunch post that hadn't shown up in my search and that I'd never noticed despite hitting HN several times a day, because they hadn't been upvoted at all and had quickly plummeted off the first few pages. I hope a solution can be found that cuts down on both redundant submissions and occupying comment space with annoying complaints about redundant submissions (and that obviates the need for annoying complaints about the annoying complaints).
I'd suggest trying just eliminating awarding karma to users, at least for submissions - just use the submissions' points for the ordering on the page. It takes way more effort to write a valuable comment than to hit submit for a link, so let comments become the sole determinant of users' karma. I think users would still submit pages they find particularly interesting for HN out of desire to share interesting things.
Would it really help with spam though? The rewards for a spammer to make it to the first, second, or even third page are probably worth the risk of negative karma.
As for duplicate articles, I'm not even really sure what the problem is. Is it duplicate articles (same url, slightly different) that clog up the system? Or is it the same subject being covered from many different sources? There have also been a few times within the last month where two stories about the same topic will be right next to each other on the front page--amazon/apple/other big industry announcements like to do this.
I expected that new accounts wouldn't be able to submit anything until they have a couple points of karma (from upvoted comments).
The problem is that the main page of the new feed frequently has a max age under an hour (44 min., currently) - if something doesn't get any upvotes by then, it falls off the page and becomes drastically less likely to get any further upvotes. It also can't be resubmitted for several months. Slowing this down would almost certainly be good - quite a few interesting posts get submitted at the wrong time and never get noticed.
We might take this concept further and make it a "wager" of karma with a minimum bet, and you get gains at some rate proportionate to the size of the bet. But extending it in this way, it would probably also have to have some effect on the front page algorithm for a large bet to ever make sense. Which in turn could lead to a stagnation where elite posters are constantly sinking thousands of karma points into "sure things" and nobody else can compete.
I think this would exacerbate the "comment quality" problem by encouraging people to make lots of sub-par comments so they can gain enough karma to submit their latest blog post.
Well, consider me happy (though weirded out a little) with the lack of points on comments - in retrospect, I think a lot of my downvoting behavior has to do with what I think a post should have earned, not what I personally think about it, and that's probably bad.
OTOH, I'm in the habit of scanning for double-digit comments when I want to save a little time, so I miss that.
Sometimes, I do the reverse of what you described. I upvote comments that are at -1 (even though I wouldn't have upvoted the comment if it hadn't been less than 1.)
In these cases, I have a neutral feeling about the comment, but upvote it because 0, -1 etc. seem unfair.
Hiding the points will mean that HN will lose some of its self-correcting capacity (when it comes to comments that were downvoted unfairly).
Anyway, it will be interesting to see the results of this experiment.
Actually, I think there is a flaw in not providing the number: You have no idea which comments are generating activity
When my number changes, I click threads and scan for a different number than I recall for my posts.
This tells me which ones are still being read.
If there was a visual indicator of posts that had the arrows ticked since I last looked - this would be solved.
Otherwise, it is akin to throwing your comments to the wind with no indicator of how each is received unless they get greyed out by down voting.
EDIT: I don't feel like I have any incentive to click the arrows on anything if I cant see the result of that click. This really really is a bad design choice.
At a minimum we should be able to see our own scores.
EDIT #2: Also, assuming two people reply to a questions asking for a recommendation for X with a link -- you have no idea what the crowd's opinion may be of the links without them replying.
Typically good content is VOTED UP as an indicator that it is, in fact, good content.
The more and more I think about not seeing content scores, the more obvious it is that this is just flat wrong to not show them.
Can no longer see even my own points. Don't think that's positive - I like the feedback over what comments I make that people like and which are less well received.
I would be okay if there were some way to tell roughly how many points a post was getting, even if it wasn't precise. For example, some sort of color coding that moved to warmer colors on a roughly logarithmic scale.
Yeah, I really miss being able to scan for high-point comments already. I've only spent about 15 minutes on HN with the new changes, and I've already basically started ignoring entire pages of comments because they'd take too long to sift through to find the insightful stuff.
What you find by scanning for high-point comments is not necessarily insightful stuff but rather, most upvoted stuff. Then you upvote it, making it feature even more prominently on the page. The comments that you didn't read, no matter how insightful, get no points from you so they fall further down the thread. I've seen this happen, joke comments positioned very high while truly good comments are very low because nobody had the patience to reach the bottom of the page, read them and vote.
Certainly. I understand what is happening. But either way, I don't have the time nor the desire to read every page of comments in its entirety, and I like having some kind of filter.
I fear I'll spend more time on HN reading every single comment. Good for the less well-known posters who don't get upvoted by reputation alone, but bad for me and my work. However, having observed and stated that fear, I can now work to counteract it should the post counts remain hidden.
This is true for replies directly to the article, but replies to the comments turn everything into a huge mess. I liked being able to see at a glance whether the thread below a comment was worth reading as compared to the next comment.
I don't know. If I can't see the little score increment or decrement in response to my up/down vote, I find I have very little urge to actually click either button. With no visual feedback to tell me "Yes, that click actually changed something", where's the satisfaction in voting?
And your noob throwaway account is now highlighted in Green. Check out the Noob Stories page for more (or to see if it's already changed back) - http://news.ycombinator.com/noobstories
I had an interesting first experience with the hidden vote count. I was reading through comments in a thread, scanning for those with high vote counts, skipping everything else. Subconsciously I saw a "1" next to every post, ignored them, and skimmed along.
Eventually I started reading some of the comments in detail, thought one was pretty good, but subconsciously still saw a "1" next to it. I didn't vote and moved on. By the third or fourth time I did this, I asked "Why isn't anything good getting votes?" After a closer look, seeing that there was no "1", it was obvious my subconscious was fooling me.
Expectedly, I'm more inclined to read a comment with a high score since it's filtered. Not surprising. But this was striking; I'm more inclined to "vote" for something if it already has a high score, regardless of its content. It's a popularity-based multiplier effect.
It would be interesting to see what happens to the "bell-curve" of vote distribution as vote count remains hidden (mean/median/mode/standard deviation). I'd predict that highly-voted comments won't be as common, and maybe lesser-voted comments might get more votes.
Has anyone else noticed or experienced this? Or maybe something else entirely?
I... I feel quite pushed away now I can't see if people are liking my contributions. I feel as if maybe people aren't liking my comments, and I can't tell so I can't adapt. I post comments similar to those I know users like, in the hope that I provide value to others, without knowing how well a comment is being received how do I provide what users want?
I would like to see point display re-enabled for the owner of the comment.
> I would like to see point display re-enabled for the owner of the comment.
I second that.
Now, that up-votes are not visible, I have to write it out. Something that makes me wonder, Will this change increase the "me-too" comments count to show agreement, which in-turn would increase the noise in comments even more?
I think the point is not to write for what users want. You should write based on what your personal opinions are. That way we get more new and interesting opinions.
I'm not sure I care to spend half an hour writing out a new and interesting opinion if no one ends up reading it. With no points visible, obvious feedback of people actually reading it is limited to reply comments.
"I can't see if people are liking my contributions. I can't tell so I can't adapt."
On the one hand it's nice to see feedback so that you know to adapt to the social norms of the site. On the other hand, I think it's great to not care so much about what others think and just focus on making your comments accurate, informative, helpful, kind and enlightened.
Is there a way we can see the current list of changes that are in production? While things like new users being colored green are obvious, other things such as a change in the decay function on the front page may be less obvious. Just curious to see what's going on at any given time.
I'm coming late to this conversation, as I was mostly internet free last week, but
a) Thanks for doing all this work
b) I imagine you generally feel this way, but I would love to have changes err on the side of keeping the community small UNTIL it proves it can scale in culture and quality.
I say this having lived through the following community site's initial quality and esprit-de-corp rise and fall:
Kuro5hin, Slashdot, Digg, Reddit
Probably the only truly excellent community I was part of which did not have this problem was the Plato Network, but I expect it died before it could grow into many of the growth/quality problems HN or any of these popular sites face.
To my mind, the idea that one is required to grow beyond one's quality and community goals need not be true. Another way to say this is that if we graphed the ability of community websites to attract new members against their ability to maintain / improve quality and culture, so far that graph is significantly below the 1:1 line.
Creating technology to change that slope above 45 degrees would be a totally huge gift to the world, seriously. On the other hand, the best sites out there might be at less than 25 degrees right now, so even a little would be a big improvement.
All that to say, I'm all for experimenting, and I know for sure that you don't really want to start HN(^2); you'd rather keep using and feeding HN in the right way -- I always find a big goal / framework to be helpful, and I haven't heard you say much about what your longterm goals are here; since talk is cheap, take mine!
One additional thought I had here; based on this graph idea, it seems to me a useful thing to do would be to create some sort of way to get a handle on community on quality using metrics.
They would probably be terribly inaccurate metrics at first, but they would also provide a means for people to decide how to assess.
As a start, you could use people, for instance paying people on Mechanical Turk to evaluate a post and its comments for specified HN guidelines; these could be used as a first attempt to get a feel for how well a given day / week goes.
I was thinking the same thing, but maybe also using sentiment analysis or other NLP techniques. Even simple word count trends might yield some insight.
If we could get a dump of all the comments tagged with submission time that would be a nice starting point for people to throw their algorithms at.
The "green user names" might be nicer if it was a continuum -- that is, "newest possible" would be bright green, and it would fade to gray over the course of days or weeks.
An observation about the disappearance of karma on comments, is that in searching older threads (I'm currently looking at a more technical thread ~1000 days old) is an inability to distinguish between the utility of comments.
E.g. if there is a solution set of A,B, or C to a particular problem or question, it's impossible to tell if 20 people thought A was a good idea while B and C were both 1 point answers.
Maybe threads past a certain date threshold could display these vote counts?
Also while you are hacking on it pg, can you implement the bookmarklet hack to collapse/expand comment trees? often it's fun to be able to just see each conversation starter and dive into one that's interesting, rather than the expanded list view which is hard to parse what are the interesting conversations.
this is the code, i don't recall where it came from:
The problem here fundamentally is that there's not a clear metric which correlates with "good", because there's no clear definition of what "good" is. (This is my utility function. There are many utility functions like it, but this one is mine.)
It's a social problem, and as such I'm intensely suspicious that any technology can fix it. In particular, I suspect that any technological solution which "works" for some people is going to amount to a dictatorship of that sort of people - which is OK! It's a big internet! Everyone else can go somewhere else! - but doesn't hit our geek pleasure centres like an elegant algorithmic solution is.
In other words: if I were trying to fix this, I'd cut straight to the chase – find and engage some people who shared my understanding of what a good HN looks like, and empower them to moderate aggressively and wield the banhammer.
Can you please consider reversing the "xx comments" and "flag" links? I think everyone here is used to clicking the last link under a submission to go to the comments!
Either reversed, or can we get the comment link up to the right end of the top line of the submission, so that I'm not forced into scanning over and navigating over the "flag" button on the way to reading the comments?
So here's a problem -- I think we all know that up/downvotes are routinely used to agree/disagree with the comments. With one notable exception, which is when the comment sits at 1, in which case it is not likely to be downvoted in disagreement.
Take parent comment for example. I disagree with it and if it had a score of 2+ I would've downvoted it, because my gut reaction was not "extremely positive". Now however I look at it and do not know if it's 1 or 2+, and so I cannot cast my disagreement vote without potentially pushing the comment into non-positive range.
(edit) Actually... what if there was separate agree/disagree indicator for each comment. As in "I was going to say something, but this comment is exactly that -> agreed". This will turn up/downvoting back to its original role of interesting/junk quantifier.
I was always under the impression that voting is not to be used for agreeing or disagreeing, but rather to reward a good contribution to the discussion (and punish a bad one). I have on numerous occasions upvoted a comment even though I disagreed with it just because I think it has been valuable to the overall discussion.
That's the idea in theory. In practice, it's clear that upvotes/downvotes are often used to guide the discussion in the direction that the voter wants it to go, which is often towards his/her innate biases.
I think this ability to vote in two different ways would be extremely interesting.
The agree/disagree voting stated above would give people insight into what others personally think about their comment and possibly push people to ask themselves, "Is this a useful, interesting comment, but I don't like it because I fundamentally disagree with it? Or is this a bad (meaning - does not contribute or against rules) comment?"
If this system worked, some of the most interesting comments could be ones that are voted up a lot, but very disagreed with.
Although looking at it from the other side, I could say it would make it harder for new users to participate/get up to speed, voting types could be confused, and it would add another step to a somewhat complicated process. (Plus people just might not like it)
edit: Fixed grammar and clarified some of my ideas
A lot of people on HN, myself included, consider votes to signal contribution to the discussion and not agreement/disagreement. Your comment is right in that they are often used as such, but I don't think it's desirable.
Yeah, I know the guidelines, but pg himself said that using voting to express (dis)agreement was OK. I don't have a link to his comment, but it was mentioned in another thread few days ago. To me his position confirms that the voting function currently sits with one butt on two chairs, and this may be the root of the fanboy-style mass-upvoting problem with comments (that, again, was pointed out by pg not few days ago).
It would be good to provide a link. I've been through the last 8 days of pg's comments which you can browse here and haven't seen anything to that effect:
Discussions will end up quite boring if people use votes to agree or disagree rather than expressing their point of view in a reply. I tend to think of votes on HN similar to the options for moderation on Slashdot: vote up for insight, vote down for off-topic or flamebait.
I have had this bookmarked for more than a year, as this issue comes up over and over and over again. Sometimes downvotes to express disagreement with the stated position in a comment is the most concise way, the way most friendly to other readers of HN, to indicate that the comment didn't add value to the community.
Of course, pg is experimenting right now to see if different software settings make upvotes, downvotes, flags, and so forth have better or worse effects on the community as a whole and on particular threads.
http://news.ycombinator.com/item?id=117171 "I think it's ok to use the up and down arrows to express agreement. Obviously the uparrows aren't only for applauding politeness, so it seems reasonable that the downarrows aren't only for booing rudeness." - pg, 1150 days ago
http://news.ycombinator.com/item?id=658691 "IIRC we first had this conversation about a month after launch. Downvotes have always been used to express disagreement. Or more precisely, a negative score has: users seem not to downvote something they disagree with if it already has a sufficiently negative score." - pg, 665 days ago
Rather than keep trying to keep educating people that you don't want them to do what apparently is natural (I think digg and reddit et al have conditioned the upvote/downvote behavior upvote things based on agreement/disagreement/amusement.)
Maybe what's needed is to try an explicitly enable both these behaviors. Something like let the arrows become "popularity/reddit-mode" and add a separate "excellent" button next to "flag". Then you have two metrics without the annoying slashdot vectors. You can then either use the up/down votes or just leave them there as placebos.
I'm probably in the minority, but I actually like the concept of the slashdot-vectors as a way to give non-verbal feedback a bit more specific than "up" or "down".
The crux is of course in the choice of adjectives. I wonder what would happen if one would let the users choose free-form tags...
On the other hand, "disagree by replying" isn't always a good thing.
For instance, deliberate trolling is far better downvoted and ignored than engaged with directly. If a troll gets twelve replies going "Actually I disagree with your Hitler-was-great comment and here's a few reasons why..." then the troll has won.
Alternatively, a truly stupid opinion, even if honestly held rather than trollishly put forth, can derail a conversation. If every time the moon landing is mentioned it results in a big long conversation about whether it was faked or not (with the "yes" side argued by one random idiot) then that's not enhancing the discussion, it's just derailing it.
There are many issues on which sensible people may disagree, and many issues on which they may not. Sensible people should be able to figure out which are which, and downvote or reply accordingly.
PS. I think this comment is great, and I'm sad that I have no way of seeing whether anyone upvotes it or not.
My gut reaction was also positive, but then I realised I miss being able to see what comments the community finds valuable. Ordering within a thread as a measure of value only works if there are multiple child comments for a particular comment.
Sometimes people post long, multi-paragraph comments. Seeing a comment like that at 2 points makes me less inclined to read it than if it were at 30 points. Not being able to see points will make skimming comments slower.
Is it possible that building on a 'broken' system isn't enough to fix the problem(s)?
The reason I point this out is that I think its possible that as a community grows the assumptions about the community become a part of the code and they aren't necessarily true anymore.
When a community is small by definition it has shared values. As the community grows it starts to fragment those shared values, but the code assumes that they still have shared values.
I am indifferent to the idea of down voting stories, but the code doesn't allow it because the assumption being made is that the stories being posted are being checked against the common community values, when in fact newer members (perhaps my self included on the newer point) see hacker news as the next technology/startup based Digg/Reddit/Slashdot, etc.
it would be cool if the score of an object in the system was a vector-like thingy in which every value is the result of operations created by people in your same account-demographic (e.g. accounts older than 3,6,9,12,15,18 months).
You would be able to filter for stuff that has the most total votes, the most votes among people in your age class, and the most people across all age classes at vote time (people who are now one year old but voted for it one year ago)
This way you could still select between stuff that is useful/interesting to everyone, such as a new $X by $Y and stuff that was interesting for people when they first read it but it's now boring since it's the tenth submission ("harry potter and the methods of rationality" comes to mind).
I contributed late to the last thread - but having a monthly "Erlang Day" (not necessarily about Erlang) would help establish a baseline for the kind of culture we want while relying completely on informal methods, i.e. no changes to code or rules of the community.
I like the no comment scores, orange dot idea (which has made a comeback).
I hope it leads to less of the mob voting mentality which seems to take hold every so often in some of the more popular threads and less of the "Sorry I downvoted you" comments that appear every so often.
On the other hand, comment scores are a very quick and easy way to see what's popular, which is still normally an informative comment. Having to wade through a lot of highly voted comments may take a bit more effort but will expose people to a greater variety of opinions.
Suggestion: the Rerun button allows veteran users to signal that an article has reached the front page before. Far from flagging the links for removal, the rerun button signifies that the content has a timeless quality.. it definitely isn’t news, but it is still worthwhile. A link that is getting both sufficient upvotes and rerun clicks is moved to the Reruns page, which would feature all manner of old chestnuts and stop them clogging up the front page.
I really don't think that duplicate content was ever a problem for HN. It's just not an issue.
The community has done a very good job of providing links to earlier discussions and has generally accepted that occasional repeats of old content are nothing to get worked up about and may even be desirable.
I agree. Though I'd like something to address duplicate submissions within a small window. Like the double front-page stories of bing's 30% market share...or the double ask-page entries on the new karma-less comments.
While I agree to it (I was gonna point it out myself), it is a (separate) problem already before the change. Showing the comment score change, is after the fact of (down)voting, after which not much can be done.
A better placed/sized voting buttons, should solve this issue to a large extent.
I was just about to ask you just that. I've noticed in this thread (http://news.ycombinator.com/item?id=2433424) that some comments don't show their points and that a few comments have an orange dot.
Intriguing. But I'm glad you're working on keeping the quality of HN up.
It seems that you can still see how many points a comment has when it's nonpositive.
Edit:
And there are other comments with negative points where the number doesn't show. And still other comments with positive score that are showing up. Consider me confused.
What about something like "netflix for comments?" Ie. your votes would endow you with a profile (a sparse vector) that then tailors HN so that you see more of what other similar users also upvote. This would also encourage people to vote more.
I originally thought in terms of one uber-editor that fed (votes to) a machine learning algorithm, but then why not allow everyone to be their own editor.
no points in comments: very interesting, a lot less bias. Unfortunately fast scanning of the best things is compromised. Probably it's still a good idea to avoid points, at least for a few days. When the comment is old enough there is no reason to don't show points again.
* Very high-rated posts could be distinguished subtly. (Maybe a small element is bolded or italicized.)
* Skimming might actually be a symptom of something wrong. Like too many low-signal posts in a thread. Also, maybe skimming even correlates with people posting too fast, or the point system leads to worse posts.
* There are other skim-markers. For example, post length and author.
* Maybe skimming using points isn't actually faster. Could people be wasting time judging whether the post's worth the points?
Interesting idea with the time T1 when comment wuld be so old that the score could be shown. But there's also existing timestamp T2 when the comment is too old to vote. How you'd suggest to relate T1 & T2?
I'm curious if something like the BCS ranking system in college football could work for online communities. The specifics don't translate, but the general idea is that you use a weighted combination of human and machine generated rankings. This can be seen as maintaining the user-driven voting system but tempered with an impartial community spirit moderator in the form of a machine learning algorithm.
How could this be applied to HN? Let's leave the standard karma/voting system as it exists, as that seems to generally work.
Next, determine the general distribution of votes per comment. This will allow for things like z-scores to be determined that can notice if a particular comment has received significantly more votes than usual.
Next, perform a machine learning algorithm on a corpus of comments. Something as straightforward as Bayesian filters can work, though self-organizing maps also have potential. This is effectively doing the same thing as spam filtering, but instead of simply flagging something as spam, it would provide its own +/- vote. The initial training would start with a baseline of existing comments, and then periodically, say once a night, be updated with recently added comments and votes. Additional information, such as the karma of the commenter, can also be incorporated.
The final ranking then, which here would be how high up on the page it appears, would be a weighted combination of user votes, z-score scaling, and machine votes.
Way to issue a challenge that all large communities struggle to solve!
So I have two thoughts that I think are in a different vein than many comments below (tried to read them all but may have missed some. Apologize for repeats)
1. Similar to other suggestions, but with a slight twist, modify the up/down votes to utilize the Net Promoter Score methodology. It has its issues but it reduces a really difficult problem to a simple question that provides a broader spectrum than y/s. Could limit the "would highly recommend" super-vote to one story per day so users would save those votes for those articles they find extremely valuable.
Actually, thinking about it. Would be cool to get a view of only stories that people have "spent" their one super-vote as that is signaling extreme importance. I think many people find many stories interesting, but would only find a few EXTREMELY interesting enough to spend their super-vote on.
2. One challenge is that HN has grown in size so much that there is no set of top stories to satisfy the entire group. Would be interesting to provide a view that matches your personal preferences. Reddit does this by subreddit, stackoverflow by tags. My personal background is personalization in the context of eCommerce, which looks more at user segments. So users who find hard-core tech knowledge interesting vs. VC news vs. geographic location. In some ways this is already being done via segmentation in the classic view: http://news.ycombinator.com/classic Are there some other obvious segments on HN?
I'm seeing green usernames for submissions and comments by new users (not sure exactly what the threshold for "newness" is, the person I saw created their account in the past hour)
I think using green usernames for 'noobstories' is wrong. Green is usually linked to positive behaviour, earned reputation/karma or a special (empowered) user status; not for indicating that a user is new.
As I can see, every user with a submission on /noobstories is green, so that's probably the threshold.
Edit: On further inspection, it appears I'm wrong, there are users there who are not green. The newest users I could find that are not green created their accounts 6 days ago. Users who created their accounts 4 days ago are green. I haven't had any luck in finding users who created their accounts exactly 5 days ago. So the threshold is either 5 or 6 days.
Here is an idea. Individual karma scores, no matter how they are scored, will create certain kinds of incentives. What if you create a global karma score for the entire site, which is an aggregate of all user activity? Posting, commenting, voting, would all not only be scored for each user individually, but for the site overall, or perhaps in the future topic categories could either be teased out by hand, or assigned using some NLP technique like LDA, and each topic could have a score.
It would at least be interesting to see the health of hacker news on a 0-1 scale somewhere on the top menu bar, and maybe some little stats behind the score to help illustrate the thought process behind the score.
You could use cosine distance for each posting from the n highest scoring posts, and take the inverse of the sum of the distances to score a potential for each post. The sum of all the potentials of all postings within a certain time period, taken as a ratio over a trailing average potential over a longer period would a an interesting way to possibly get a view into how diverse and how high quality each posting tends to be.
A sitewide score would also allow you to make each user's score be a weighted combination of individual score and their contribution to the whole. Consider it a form of shared fitness.
This is supposed to be hacker news, if you don't like the comment counts, hide them yourself. Hack it into something that you do find easier to manage. I started writing my own chrome plugin a year ago and add to it as needed (voting ppl I like up, others down, etc.)
for me HN is perfect - you will never make it perfect for tens of thousands of people by changing it on the server.
If you start accepting user submitted tweaks now it will probably never end
One thing I've noticed about the way I use HN, is that I find I keep going to my page, then my comments/posts to see if there are any responses, so that I can decide to respond or not.
I think it would be nice to have a number next to my score, showing me that their are responses, and maybe link that right to my comments/posts page. Would save me wasting time looking at those pages when there hasn't been any activity.
I support the decision to downplay karma within comments; I would even experiment with removing that data from the article listings view. This is a data driven group of people who are simultaneously very competitive. It's likely that the majority of readers aren't maliciously competitive regarding karma, but this competitive, data driven nature likely has a contributing factor to some of the referenced issues.
If you notice, many of the comments themselves are about the external of the karma system, such as justification for downvotes, using karma as a currency, etc. The current iteration, using karma on the back end for page ranking purposes, will likely be the best long term solution. I will know the community's response to my own comments, but I won't be able to notice that a HN "super member" received 20+ votes and try to emulate their content. Over time, comment and article rank, the original idea behind a voting article/comment system, will become the goal, rather than earning 2 upvotes.
To put this in perspective, do a Ctrl+F for "karma" and "vote", then do the same for "quality."
My solution to the voting problem is to separate out good posting behaviour and good voting behaviour.
Weight votes by reputation, calculate reputation by judging each vote for difficulty (did you vote before it had obviously won / lost) and success.
So you get reputation for voting up at -10 when it ends up at 1000. You lose reputation for voting up at 1000 when it ends at -10.
Since judgement is subjective you need a bunch of moderators to seed the site with good judgement. Then so long as they over-ride votes on the most obvious examples it should be a self correcting reputation system.
I don't understand why in HN's current system voting power comes from karma which comes from occasional good posting. You could spend most of your time flaming, getting into arguments and down voting everyone you disagree with but so long as you're quick on posting PG's articles or write the occasional good comment HN views you as a model citizen.
Be interested in any critique of this as it's how ISDaily my news start-up works. (http://www.isdaily.com)
The red dot helps quickly identify comment threads that might be compelling (something I used score for before).. but the lack of scores on the comments view (using the 'comments' header link) makes that view almost useless to me.. I don't use it all that often, but occasionally peruse it to identify threads that may be more interesting than their titles led me to believe.
On the front page, can you switch the order of 'flag' and 'N comments' - I think it's a 'Fits Law' type thing. I click into the comments for basically every story, but having to get between the time stamp and the "| flag" indicator ads a second of hesitation to make sure I'm not accidentally flagging something that I want to read.
Not showing scores is curious.
For one — and this isn’t a criticism — it makes voting feel slightly idempotent.
For another, given the way both arrows always disappear after a vote, it makes it tougher to see if you clicked the “correct” arrow (always been an issue given their size and proximity).
Along with hiding comment vote counts, maybe you could try experimenting with hiding story vote counts as well. It may encourage readers to comment on stories they are genuinely interested in rather than the top voted stories of the day.
Thoughts re: zero karma display currently in place:
I'm getting a lot more up-votes for what I consider to be relatively mediocre comments. Some display is useful, as it helps prevent such artificial inflation. It is interesting in that it does mask whether the top comment is high-rated or merely new - I've gotten used to it with my Flattehn extension and I sort of like it, it helps spread the votes around a bit more.
Not sure what all needs to be done to counter-act everything. Keep up the experimenting! And may I request tool-tips on color / display differences, like the green names had a little while ago? Or a current-version legend somewhere?
You might just consider making the flagging ability more apparent, and give it to some of the more senior members of HN. I only recently discovered it, but it has been useful in removing posts that belong on Reddit.
One of the problems of such communities is the echo chamber nature of hearing the same voices. I'd be interested to see what happens if you made it so that the more karma points you had, the harder it is for your comment to rise to the top. If a post by a leaderboard member is really interesting, then it'll get upvoted a lot- but because of the group think aspect leaderboard members tend to get upvoted a lot because they're in the leaderboard.
nb: This problem may be mitigated by not having scores visible.
This way, all comments/posts that generate some goodwill are rewarded, and the incentive not to post for fear of hurting the average is removed.
And the log will frustrate karma whores b/c they won't be able to separate themselves from the pack as easily.
Also I think some measure of how controversial a comment/post is would be useful, since it would help find interesting areas where the community is in disagreement.
Yeah, usually I don't have time to read all the comments, so I just jump to the most popular ones and read those. Now I can't see what the most popular comments are.
Trouble is, I often (used to) read highly-voted comments that were children of poorly-voted branches. I don't have the time to read every comment, so I find I'm reading a lot fewer comments now.
On the other hand, I'm now less emotionally invested in replying to comments to my own comments. Before, if replies to my comments got more votes than my own comment, I might be inclined to argue my case further. I'm less likely to do that now.
I'd love if you tried the shipped idea suggested by ericb[1]. A simple input-form in the profile page that someone could put a link to and that would light up a 1x1 pixel next to their username if they've shipped a startup yet.
I love not seeing karma on comments. I've realized that I've made three comments today, where I rarely comment normally (once a week or less). I don't know why, but it takes away the "you'd better write two paragraphs and predict the group-think" feeling that stops me from contributing more.
I think the concept of seeing your own karma total goes hand in hand with seeing karma scores on your own comments. People will want to track and monitor their karma if they are able to.
FWIW, I'm already liking not seeing the score of comments. I can see how it really contributes to the discussion. Maybe do a poll later to vote for which changes should stick around?
My experience: I just read the comment thread on an HN post and without the points beside each comment my brain acted like a crutch had been removed. I approve :)
I think it's good that the usernames don't show. Other HN members have expressed a similar opinion over the years... it prevents voting for/against the person keeping your impression of the comment (at least the first impression) based on the content and not the author.
I strongly disagree. It's very useful to be able to qualify comments with what I know of that user, especially when I don't know much about the field being talked about.
I'm much more likely to take tptacek's advice on security more seriously than someone else's. Likewise with grellas's legal analysis, Patio11's small business/SEO advice, etc. That their names along might cause high karma for them isn't a bad thing. Quite the opposite, it causes their comments to be pushed to the top, which is almost always a good thing, as they're consistently the best comments in those threads.
It gives some measure of ability to gauge trustworthiness.
But I'm saying that the first impulse impression/reaction for a comment should be based on what it says, not who's doing the saying. After you read it, you can qualify it by its author. It's not like I'm suggesting we hide the author or anything - it's right there, when you need to find out who wrote what. But don't make it stick out any more than it already does. Put the content front and center, and make it the focus of attention. After you've read the comment and judged it for what it says and how its written, then reflect on who wrote it and what kind of weight that will give it.
Hacker News' display of usernames borders on anonymity; it's not highlighted nor emphasized in any way. It's like HN is an invitation to debate with its hivemind.
It should at the very least be easier to scan the usernames when reading large threads. Now, it's either a strain or something you forgo entirely. I'd be interested to see an eye movement analysis of HN users in relation to the username displays.
Grudges are what Ignore features are for. (And the opposite for Buddy systems.)
I recently wounded up criticizing a start-up pitch in a thread, and it turned out that the person I had the conversation with was one of the founders. At least the OP - and preferably the coworkers - should be highlighted. The latter could be achieved by letting the OP include user names, when s/he creates a thread.
Individual comments sans karma-per-comment is superbly smart. Seems like it will encourage higher quality comment quality across the board, as a function of less-biased individual comments in aggregate.
Another + side-effect: decreasing the number of attack down-votes. It is annoying to see scores < 0 due to another person (coward) who uses downvote to "disagree" w/a valid point, and then everybody jumps on the bandwagon. It is possible to disagree without downvoting; just takes a little more thought / energy.
First up, this is nice. The absence of comment scores throw people around a bit, and people are generally more careful since it is difficult to directly receive feedback.
Secondly, the madness associated with karma is gone. Can you imagine in post-earthquake Japan where for a brief few days, it didn't matter whether you are a CEO or have a million dollars. It is a brief moment where we are reduced to who we already are: humans. Beyond the profit motive. I wonder if this is sustainable in the long run. It'd be like Communism (?) gulp.
Perhaps there could be karma-free days as an experiment where for those days, you couldn't earn karma even if you tried. Kind of like Sabbath. It forces a different perspective.
I hope you don't turn the points display back on. Hiding points makes it more likely that users will actually read a comment before voting it up or down.
Please fix downvoting. Even right now, there's a thread that I can my post be downvoted, and then my reply later that sets the record straight be upvoted. Apparently people here vote before informing themselves or reading the rest of the thread. Regardless, not being able to change a vote, especially if a simple misclick is made, is very, very frustrating.
Okay, I'll come clean. I've been experimenting with HN for over three years.
At first, I would try out HN and the effect on me was quite mild. I simply didn't see what the fuss was about. But the most I experimented with HN, the more I realized how powerful it is.
Nowadays, I use HN daily, sometimes more than once daily.
Sure my relationships with people who don't use HN have suffered. But who cares? They don't understand. That's why they don't use HN.
Neat - even though we can't see votes, downvoted comments are gray. Maybe all of a thread's comment's color will be normalized between Black and Invisible, depending on a comment's relative number of upvotes (or negative score).
This intrigues me. I don't get your comment. I thought it was going to be a joke about drugs or some church or something. Is this in response to the title 'Experimenting with HN'?
Having more articles on the front page that aren't based on a scramble for duplicate-submission karma would probably improve the overall discussion threads, too; those posts tend to draw a lot of shallow comments.
It would also help with spam. Win win win.
I also wonder what percentage of upvotes for submissions comes just from duplicate submissions - maybe those should be counted differently (or not at all)? If the front page is already full of threads about some news about Apple (or whatever), being first to submit a redundant (but distinct) post is disproportionately rewarded, yet reduces the signal/noise ratio even further.